Northern Prairie Wildlife Research Center
Regression models demonstrated the impact of agricultural development on breeding duck populations in the PPR. PERCROP and PERGRASS were included in most models, as suggested by the exploratory ANOVA tests, but models were inconsistent among years and species. These results suggest there are landscape effects on waterfowl breeding pairs, but the relations are more complex than we were able to examine here. Models using numbers of pairs or γ may be improved with the incorporation of other factors not considered in this study, such as habitat fragmentation.
PERCROP significantly affected total number of duck pairs as anticipated (more ducks were present when PERCROP was low) and contributed to the best regression models. In 1996, this relation was complicated by an interaction with TOTDRAIN. The response of individual species was more complex and, except for blue-winged teal, included interactions with PCTNMOD and TOTDRAIN. Cowardin and Sklebar (1997) noted species differences but did not conduct tests to compare among species. Differences in species biology and habitat selection contribute to these differences among species and the greater complexity in response. The lack of a significant effect of PERCROP for pintails may reflect the greater use of cropland by nesting pintails (Greenwood et al., 1995) or be confounded by the east-west trend noted above. Annual differences in significant effects within species makes it difficult to explain interactions and reduces our confidence in using species-specific measures as good indicators of landscape condition. A longer study may clarify what other factors are contributing to these patterns or the nature of the relations among these variables.
We included γ as a response variable in our analyses as an alternative measure of duck response to landscape conditions. Cowardin et al. (1995) developed baseline regression equations for the five duck species to predict breeding pairs as a function of pond size, based on data collected in the Arrowwood Wetland Management District under moderate water conditions. The correction factor, γ (the ratio of observed to predicted number of ducks), allows for adjustment of these original regressions for each area and year. Because 'predicted' breeding pairs is based on ponds by size in the study area that year, γ effectively adjusts for annual or area-related differences in the response of duck pairs to ponded wetlands by area. Where γ is <1, as found consistently in the Red River Valley and many Drift Plain study areas, fewer ducks are present in an area than predicted given those wetland conditions. This would suggest other factors are contributing to the relative lack of ducks, such as landscape factors or wetland regime. The magnitude of γ therefore may provide a useful measure of the landscape condition for ducks, given those wetland conditions. However, of the landscape variables examined, only PERGRASS showed any significant effect, and the results were not consistent among species or years. The curvilinear relation between mallard pairs and pond area varies with wetland regime (Cowardin et al., 1988a), suggesting that better results might be obtained if γ specific to seasonal wetlands were used rather than γ's from all wetland regimes combined, as used here.
Under the systematic sampling scheme used here, the distribution of landscape variables tended to be clumped toward one end of the range. For example, the median of PERCROP in this study was 73.2%, and only four study areas had <33% of upland habitat in cropland. In comparison, the Prairie Pothole Pilot Study I used study areas specifically selected to provide two non-overlapping extremes: the median value of cropland in cropland-dominated areas was 91.5% whereas that for grassland-dominated areas was 23.9% (Cowardin and Sklebar, 1997). Although it would have been desirable to have better representation of grassland-dominated areas in this study, the systematic sampling results reflect the dominance of annual-crop agriculture in the entire glaciated region of North Dakota. To better examine the relation between landscape variables and duck numbers, we need to select areas which would provide a more uniform distribution of the landscape variables of interest. The split of landscape variables into high versus low values in our analyses was artificial but allowed exploratory analyses of relations between estimated numbers of pairs and the selected explanatory variables.
Before the results of this study are used for monitoring landscape condition, several caveats should be considered. First, the study was conducted during two wet years, when number of basins holding water and basin size were high compared to the average. In 1996, the Palmer Hydrological Drought Index averaged 4 (on a scale of -8 to +8, 0 being normal) (National Climate Data, 1998), and May pond numbers in the north-central U.S. in 1995 and 1996 were at record highs (U.S. Fish and Wildlife Service, 1997). We anticipate that the relations found here would differ under different water conditions, particularly in dry years. Other waterfowl studies that have categorized years as wet, moderate, or dry have found breeding duck responses differ relative to water conditions (e.g., Sorenson, 1991; Serie et al., 1992; Anderson et al., 1997). Second, numbers of breeding pairs available to settle in an area can be constrained by the population available. Johnson (1996) found numbers of blue-winged teal and pintails in a North Dakota area were correlated to their continental breeding population. Populations of prairie-breeding ducks change over time and, for some species, recent population changes have been dramatic. Gadwall numbers increased >200% and pintail numbers declined by 40% over the past 10 yr (U.S. Fish and Wildlife Service, 1997). Such changes likely influence the numbers of pairs present annually on specific areas. Comparisons of estimated duck numbers among areas within a single year would not be affected by this, but it would be a factor in temporal trend analyses. Our understanding of how changes in the continental population influence local numbers, and at what scale such an effect can be detected, is poor and those relationships likely vary with species.
Naugle et al. (1999) demonstrated that life history characteristics also should be taken into account when considering wetland species as indicators of local or landscape-level conditions. Black terns (Chlidonias niger), a mobile species that forages in multiple wetlands, had smaller area requirements in heterogeneous than in homogeneous landscapes and were more likely to occur in landscapes where grasslands had not been tilled for agricultural production. In contrast, the occurrences of two sedentary species that forage largely or entirely within their nesting wetlands (pied-billed grebe (Podilymbus podiceps) and yellow-headed blackbird (Xanthocephalus xanthocephalus)), were explained solely by within-patch variation in wetland measures. Average home ranges and habitat use patterns for ducks also vary among species due to differing life-history characteristics. The scale of measurements likely will affect the relationships between duck occurrence and landscape measures, but this issue has not been adequately addressed. Biologists need to consider the significance of life-history characteristics, and the scale at which those characteristics are expressed, when selecting species to include in studies monitoring wetland conditions. Multispecies approaches should consider including species representing a range of characteristics appropriate to the scale of interest.
The intent of this study was to test whether waterfowl breeding populations could be used as a biotic indicator for monitoring the status and long-term trends of ecological resources in the region (Peterson, 1994). Resources needed to monitor either landscape features or waterfowl, following methods used here, include acquisition of aerial photographs each spring and extensive GIS processing; waterfowl monitoring also requires support of field personnel during May. Although it may appear to be more cost-effective to simply monitor the percentage of cropland in upland habitat, we estimate GIS processing effort for water data, needed to extrapolate duck counts from the roadside transect surveys, is approximately 5-6 times lower than the effort needed to determine the percentage of cropland in upland habitat. This savings more than outweighed the costs associated with field work. Thus using waterfowl as an indicator remains a more cost-effective approach.
In conclusion, our results suggest there are landscape effects on waterfowl breeding pairs but the relations are complex. Our results are confounded by the geographic differences in upland habitat which are largely the result of geologic and climatic factors. Percentage of cropland in upland habitat, originally considered a proxy to wetland habitat condition, does have the most consistent effect on the waterfowl measures examined, but this variable is among the most strongly influenced by geographic location. Most of the potential biotic indicators of wetland condition examined here would be appropriate for temporal trend analyses, but because of inherent geographic variability would not be appropriate for single-year geographic trend analyses without more extensive evaluations to improve explanatory models.